Invoice Reader API - FastAPI

FastAPI reimplementation of invoice-reader-api, originally built with Flask.

REST API that extracts structured data from PDF invoices using FastAPI and OpenAI GPT-4o-mini. Returns validated JSON responses with automatic Swagger documentation.

Features

• Upload any PDF invoice and extract structured data automatically
• Handles invoices of any format or layout using GPT-4o-mini
• Returns vendor, client, line items, totals, tx, and currency
• Automatic request/response validation with Pydantic
• Interactive API documentation at /docs
• Tested with real invoices accross NZD, AUD, and USD

Endpoints

Method	Endpoint	Description
GET	/api/health	Health check
POST	/api/invoices/upload	Upload and process a PDF invoice
GET	/api/invoices/<id>	Get a processed invoice by ID
GET	/api/invoices	List all processed invoices

How to use

1. Clone this repository
2. Install dependencies:

pip install -r requirements.txt

3. Create a .env file:

OPENAI_API_KEY=your_key_here

4. Run the API:

uvicorn main:app --reload

5. API will be available at http://127.0.0.1:8000
6. Interactive docs at http://127.0.0.1:8000/docs

Example requests

curl http://127.0.0.1:800/api/invoices/upload \
  -F "file=@invoice.pdf"

curl http://127.0.0.1:8000/api/invoices

Tech stack

• Python 3
• FastAPI
• Uvicorn
• Pydantic
• pdfplumber
• OpenAI API (gpt-4o-mini)

Notes

• AI parsing over regex: Invoice formats vary too much for regex to be reliable. GPT-4o-mini handles any layout consistently.
• temperature=0: Deterministic output ensures the same invoice always returns the same result.
• Pydantic models: All request and response data is validated automatically — no manual type checking needed.
• In-memory storage: Intentional for simplicity. Production version would use SQLite or PostgreSQL.